Re: VIRTIO_NET_F_MTU not negotiated

["Michael S. Tsirkin" <mst@redhat.com>]

On Thu, Jul 28, 2022 at 05:51:32AM +0000, Eli Cohen wrote:
> > From: Michael S. Tsirkin <mst@redhat.com>
> > Sent: Wednesday, July 27, 2022 6:44 PM
> > To: Eli Cohen <elic@nvidia.com>
> > Cc: Eugenio Perez Martin <eperezma@redhat.com>; qemu-devel@nongnu.org; Jason Wang <jasowang@redhat.com>;
> > virtualization@lists.linux-foundation.org
> > Subject: Re: VIRTIO_NET_F_MTU not negotiated
> > 
> > On Wed, Jul 27, 2022 at 10:16:19AM +0000, Eli Cohen wrote:
> > > > -----Original Message-----
> > > > From: Michael S. Tsirkin <mst@redhat.com>
> > > > Sent: Wednesday, July 27, 2022 12:35 PM
> > > > To: Eli Cohen <elic@nvidia.com>
> > > > Cc: Eugenio Perez Martin <eperezma@redhat.com>; qemu-devel@nongnu.org; Jason Wang <jasowang@redhat.com>;
> > > > virtualization@lists.linux-foundation.org
> > > > Subject: Re: VIRTIO_NET_F_MTU not negotiated
> > > >
> > > > On Wed, Jul 27, 2022 at 09:04:47AM +0000, Eli Cohen wrote:
> > > > > > -----Original Message-----
> > > > > > From: Michael S. Tsirkin <mst@redhat.com>
> > > > > > Sent: Wednesday, July 27, 2022 10:25 AM
> > > > > > To: Eli Cohen <elic@nvidia.com>
> > > > > > Cc: Eugenio Perez Martin <eperezma@redhat.com>; qemu-devel@nongnu.org; Jason Wang <jasowang@redhat.com>;
> > > > > > virtualization@lists.linux-foundation.org
> > > > > > Subject: Re: VIRTIO_NET_F_MTU not negotiated
> > > > > >
> > > > > > On Wed, Jul 27, 2022 at 06:51:56AM +0000, Eli Cohen wrote:
> > > > > > > I found out that the reason why I could not enforce the mtu stems from the fact that I did not configure max mtu for the net
> > > > device
> > > > > > (e.g. through libvirt <mtu size="9000"/>).
> > > > > > > Libvirt does not allow this configuration for vdpa devices and probably for a reason. The vdpa backend driver has the
> > freedom
> > > > to do
> > > > > > it using its copy of virtio_net_config.
> > > > > > >
> > > > > > > The code in qemu that is responsible to allow to consider the device MTU restriction is here:
> > > > > > >
> > > > > > > static void virtio_net_device_realize(DeviceState *dev, Error **errp)
> > > > > > > {
> > > > > > >     VirtIODevice *vdev = VIRTIO_DEVICE(dev);
> > > > > > >     VirtIONet *n = VIRTIO_NET(dev);
> > > > > > >     NetClientState *nc;
> > > > > > >     int i;
> > > > > > >
> > > > > > >     if (n->net_conf.mtu) {
> > > > > > >         n->host_features |= (1ULL << VIRTIO_NET_F_MTU);
> > > > > > >     }
> > > > > > >
> > > > > > > The above code can be interpreted as follows:
> > > > > > > if the command line arguments of qemu indicates that mtu should be limited, then we would read this mtu limitation from
> > the
> > > > > > device (that actual value is ignored).
> > > > > > >
> > > > > > > I worked around this limitation by unconditionally setting VIRTIO_NET_F_MTU in the host features. As said, it only indicates
> > > > that
> > > > > > we should read the actual limitation for the device.
> > > > > > >
> > > > > > > If this makes sense I can send a patch to fix this.
> > > > > >
> > > > > > Well it will then either have to be for vdpa only, or have
> > > > > > compat machinery to avoid breaking migration.
> > > > > >
> > > > >
> > > > > How about this one:
> > > > >
> > > > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > > > > index 1067e72b3975..e464e4645c79 100644
> > > > > --- a/hw/net/virtio-net.c
> > > > > +++ b/hw/net/virtio-net.c
> > > > > @@ -3188,6 +3188,7 @@ static void virtio_net_guest_notifier_mask(VirtIODevice *vdev, int idx,
> > > > >  static void virtio_net_set_config_size(VirtIONet *n, uint64_t host_features)
> > > > >  {
> > > > >      virtio_add_feature(&host_features, VIRTIO_NET_F_MAC);
> > > > > +    virtio_add_feature(&host_features, VIRTIO_NET_F_MTU);
> > > > >
> > > > >      n->config_size = virtio_feature_get_config_size(feature_sizes,
> > > > >                                                      host_features);
> > > >
> > > > Seems to increase config size unconditionally?
> > >
> > > Right but you pay for reading two more bytes. Is that such a high price to pay?
> > 
> > 
> > That's not a performance question. The issue compatibility, size
> > should not change for a given machine type.
> > 
> 
> Did you mean it should not change for virtio_net pci devices?

yes

> Can't management controlling the live migration process take care of this?

Management does what it always did which is set flags consistently.
If we tweak them with virtio_add_feature it can do nothing.

> > 
> > > >
> > > > > @@ -3512,6 +3513,7 @@ static void virtio_net_device_realize(DeviceState *dev, Error **errp)
> > > > >
> > > > >     if (nc->peer && nc->peer->info->type == NET_CLIENT_DRIVER_VHOST_VDPA) {
> > > > >          struct virtio_net_config netcfg = {};
> > > > > +        n->host_features |= (1ULL << VIRTIO_NET_F_MTU);
> > > > >          memcpy(&netcfg.mac, &n->nic_conf.macaddr, ETH_ALEN);
> > > > >          vhost_net_set_config(get_vhost_net(nc->peer),
> > > > >              (uint8_t *)&netcfg, 0, ETH_ALEN, VHOST_SET_CONFIG_TYPE_MASTER);
> > > >
> > > > And the point is vdpa does not support migration anyway ATM, right?
> > > >
> > >
> > > I don't see how this can affect vdpa live migration. Am I missing something?
> > 
> > config size affects things like pci BAR size. This must not change
> > during migration.
> > 
> 
> Why should this change during live migration?

Simply put features need to match on both ends.

> > --
> > MST